Hands-On Machine Learning with C++

Density estimation approach (multivariate Gaussian distribution) for anomaly detection

Let's assume we have some samples in a dataset and that they are labeled and normally distributed (Gaussian distribution). In such a case, we can use distribution properties to detect anomalies. Let's assume that the function gives us the probability of a sample being normal. A high probability corresponds to a regular sample, while a low probability corresponds to an anomaly. We can, therefore, choose thresholds to distinguish between regular values and ...